World Academy of Science, Engineering and Technology 22 2008
Choosing R-tree or Quadtree Spatial Data
Indexing in One Oracle Spatial Database
System to Make Faster Showing Geographical
Map in Mobile Geographical Information
System Technology
Maruto Masserie Sardadi, Mohd Shafry bin Mohd Rahim, Zahabidin Jupri, and Daut bin Daman
Abstract—The latest Geographic Information System (GIS)
technology makes it possible to administer the spatial components of
daily “business object,” in the corporate database, and apply suitable
geographic analysis efficiently in a desktop-focused application. We
can use wireless internet technology for transfer process in spatial
data from server to client or vice versa. However, the problem in
wireless Internet is system bottlenecks that can make the process of
transferring data not efficient. The reason is large amount of spatial
data. Optimization in the process of transferring and retrieving data,
however, is an essential issue that must be considered. Appropriate
decision to choose between R-tree and Quadtree spatial data indexing
method can optimize the process. With the rapid proliferation of
these databases in the past decade, extensive research has been
conducted on the design of efficient data structures to enable fast
spatial searching. Commercial database vendors like Oracle have also
started implementing these spatial indexing to cater to the large and
diverse GIS. This paper focuses on the decisions to choose R-tree
and quadtree spatial indexing using Oracle spatial database in mobile
GIS application. From our research condition, the result of using
Quadtree and R-tree spatial data indexing method in one single
spatial database can save the time until 42.5%.
Keywords—Indexing, Mobile GIS, MapViewer, Oracle Spatial
Database.
I. INTRODUCTION
M OBILE GIS is the extension of a geographic
information system (GIS) which has been developed
then can be run in the field. Previously GIS can only run in
the office with a desktop GIS. With mobile GIS, the user can
retrieve, transfer, update, manipulate, analyze and display
geographic information everywhere and every time. The
standard technologies that integrate in Mobile GIS application
are wireless network for transfer and access data, mobile
devices for run GIS application in everywhere, and GPS
(Global Positioning System) to detect the location.
Authors are with Faculty of Computer Science and Information System,
University Technology of Malaysia, 81310 Skudai, Malaysia (e-mails:
muserie@gmail.com, shafry@utm.my, zaha@utm.my, daut@utm.my).
249
Over the past years, Mobile GIS applications and
technological trends have a rapid development. The success
and emergence of the WWW (World Wide Web) and internet
also support this rapid advancement. The geo-information
technology had emergence from mainframe computers to
stand alone desktop computer GIS, local networking GIS, web
GIS and now Mobile GIS where geographical map and the
information are run and displayed on small mobile devices
like Mobile Phones and Personal Digital Assistants (PDA) [1].
The internet as a telecommunication network over the world
can be used for transferring geographic data. The integration
of the internet and GIS technologies as we call internet GIS
can support the demand for geographic data access and
transfer [2]. Additionally, internet can be used for
communication data between client and server. With Mobile
GIS, field data collection can be settled and send the data to
the server for further processing. Furthermore, client can
access the information needed for enhance the collection of
the geographical data [3].
Rapid improvement in the Mobile GIS technology can solve
mobile handheld devices problem like their small bandwidth,
limitation of applications capability, color resolution, and
small screen display [4]. The recent developments of internet
and Mobile GIS technology make it possible the process of
spatial data transferring, collection, processing and
dissemination with large amount of geographic data [5].
Previously, data collection and editing process in the field
take a lot of time and susceptible result some errors. GIS user
must visit to the field for take the geographical data in the
structure of paper maps. Then they carry out field edit using
draft and notes on paper maps and structure. After they got the
geographical data, the processes of field edits in the office
must be interpreted and manually entered into spatial
database. Consequently, the geographical data rarely accurate
and up-to-date as it could have been.
With the rapid Mobile GIS developments now a day, GIS
can be taken into the field using some devices like compact
digital maps, laptop (mobile computer), PDA (Personal
Digital Assistant), etc. This new GIS technology enables
company or GIS user to get real time geographic information,
update data much faster to their database and applications,
 World Academy of Science, Engineering and Technology 22 2008

efficiently analysis, display geographic data, and making R1 R2

decisions in the field.
The major concern of most wireless technology is the
overloading servers because of the system bottleneck
R3 R4 R5 R6
phenomenon. Mobile GIS applications use a wireless
technology for process transferring and retrieval data, and
then it needs concentrating to explore and solve that problem.
The system bottleneck occurs in four main areas: database, A I E 3 B C D 2 F G
network, application server and web server. The major
problem for system bottleneck phenomenon originates from
the database [6]. In mobile computing, considerable research B

has been focused on the infrastructure and architecture design A

R3
C
to try and solve system bottlenecks, such as the development I R5
of third generation mobile systems or to resolve cache
management of the database. Thus optimizing the existing
system such as fine tuning databases is an important step to D

enhance the overall performance.

To solve the problem, making some optimization in spatial
R6 2 F
database system is one of the approaches. One of the ways to
optimize process of transferring and retrieving data in spatial
R2
database system over mobile GIS network is spatial data G
indexing approach. The best spatial data indexing method are
R-tree and Quadtree. Each of those methods have different
advantages and disadvantages base on the requirement of the
applications and type of data; if Quadtree and R-tree will be E

used together for index the data in one single spatial database R1 R4
3

system then the database can be optimized with the

appropriate spatial data indexing method and can be
contributed to improve spatial data transferring speed. Fig. 1 R-tree Algorithm Schema [7]

Many researchers explored about R-tree spatial indexing,

II. R-TREE SPATIAL DATA INDEXING OVERVIEW
R-trees are tree data structures that derived from B-trees. It
is used for representing multidimensional point data, but is
used for spatial process methods for indexing multi-
dimensional information. Each node in the tree communicates
to the smallest d-dimensional rectangle that surrounds its child
nodes. The leaf nodes include cursor to the real geometric
entity in the database, as an alternative of child. The entity is
represented by the smallest associated rectangle in which they
are enclosed [7].
There are many possible approaches to divide a node. The
first goal is to allocate the records among the nodes so that
probability that the nodes will be visited to subsequent
searches will be reduced. The total areas of the wrapping
rectangles for the nodes should be minimized to achieve this
goal. Another goal is to decrease the probability that both
nodes are observe in successive searches. The area common to
both nodes should be minimized to achieve this goal. This
following figure is to describe R-tree collection of rectangles
and the spatial extension of the enveloping rectangles.
because R-tree is one of the best spatial indexing and
recommended in Oracle spatial database. Here, we discuss
some of the researchers that already used the advantages of R-
tree for the spatial indexing.
A new spatial cluster grouping algorithm and R-tree
insertion algorithm has been proposed. It introduces the k-
means clustering method and employs the 3D overlap volume,
3D coverage volume and the minimum bounding box shape
value of nodes as the integrative grouping criteria [8]. A
scalable technique called Seeded Clustering that allows us to
maintain R-tree indices by bulk insertion while keeping pace
with high data arrival rates has also been proposed [9]. A
bottom-up update strategy for R-trees that generalizes existing
update techniques and aims to improve update performance
has also been proposed [10].
A novel bulk insertion technique for R-Trees using Oracle
10g that is fast and does not compromise on the quality of the
resulting already presented [11]. A generalization for the
family of R-trees, called the Multi-scale R-tree, that allows
efficient retrieval of geometric objects at different levels of
detail has also been proposed [12].
The real-time mobile GIS based on the HBR-tree to manage
mass of location data efficiently have been explored [13]. A
Technique combining existing Q + R-tree and QuadTree in
terms of range query execution time by a high order of
magnitude has also been proposed [14]. An efficient protocol
for the kNN search on a broadcast R-tree, which is a popular

250
 World Academy of Science, Engineering and Technology 22 2008
multi-dimensional index tree, in a wireless broadcast
environment in terms of latency and tuning time as well as
memory usage has also been proposed [15].
III. QUADTREE SPATIAL DATA INDEXING OVERVIEW
A quadtree is a tree data structure in which each internal
node has up to four children. Quadtrees are most often used to
partition a two dimensional space by recursively subdividing
it into four quadrants or regions. The regions may be square or
rectangular, or may have arbitrary shapes. A similar
partitioning is also known as a Q-tree. Quadtrees has two
features which are: (1) they decompose space into adaptable
cells and (2) each cell (or bucket) has a maximum capacity.
When maximum capacity is reached, the bucket splits, and the
tree directory follows the spatial decomposition of the
Quadtree [16].
To cover the geometry, Quadtree spatial data indexing uses
fixed-size tiles controlled by tile resolution. Each tile has a
fixed-size and shape, because the coordinate space has been
decomposed a specific number of times. Then the tessellation
will be terminated. However all of this process could be
happened if the resolution is the sole controlling factor [17].
As we can see in Fig. 2, the tiles can be sorted linearly and
systematically at the exact level. Then it will visit the tiles
which determined by a space-filling curve. The unique
numeric identifiers are also assigned to the tiles.
Fig. 2 Quadtree Algorithm Schema
Many researchers had explored about quadtree spatial
indexing, even though quadtree is not recommended to use for
particular spatial indexing compare to R-tree spatial data
indexing. However, there are many advantages using quadtree
spatial indexing in the special situation. An algorithm based
on applying eigenspace methods has been presented to a
quadtree of related set images to solve the pose estimation
problem in the presence of occlusion and/or background
251
clutter. The inability to easily locate the desired object and
apply the appropriate normalizations, are efficiently overcome
by the recursive quadtree procedure [18]. Fluid flow solver
based on unstructured quadtree / octree has been explored
also. This indexing type Eulerian grids is coupled with a
spectral Finite Element (p-FEM) structural mechanics solver
based on a Lagrangian description can predict bidirectional
fluid-structure interaction (FSI) [19]. The new structure of
Multiversion Linear Quadtree (MVLQ) has been present
based on spatio-temporal access method. In this indexing
structure can be used as an index mechanism for storing and
accessing evolving raster images [20].
A novel fractal image coding based on Quadtree partition of
the adaptive threshold value has been proposed to show better
performance including the improved quality of the decoded
image, shorter compression time and higher compression ratio
with another spatial indexing structure [21]. An alternative
non-pointer quadtree node codification to manage
geographical spatial data also has been presented. In this
structure, new codification is based on a variable sequence of
z-ordered base four digits [22]. A new quadtree-based
decomposition of a polygon possibly with holes also has been
presented. An approximate this indexing structure is good for
ray shooting in the average case as defined by Aronov and
Fortune [23].
MPI collective algorithm selection process and explore the
applicability of the quadtree encoding method has been
proposed. This method constructs quadtrees with different
properties from the measured algorithm performance data and
analyzes the quality and performance of decision functions
generated from these trees. Selecting the close-to-optimal
collective algorithm based on the parameters of the collective
call at run time is an important step in achieving good
performance of MPI applications [24]. The matrix quadtree
also has been presented based on the weighted in the form of
modification. The technique of distance transform is extended
to the weighted regions and applied to the robot path planning
[25]. A new data dissemination protocol that exploits
“Quadtree-based network space partitioning” to provide more
efficient routing among multiple mobile stimuli and sink
nodes also has been proposed. Because of a common
hierarchy of cluster-head nodes is constructed where the data
delivery to mobile sinks is independent of the current position
of mobile stimuli [26]. A 3D mesh is generated by using the
original quadtree triangulation also algorithm has been
proposed with a nontrivial algorithm for quadtree
triangulation. The aim of the study was to increase the
accuracy of a terrain triangulation while maintaining or
reducing the number of triangles [27].
A new data dissemination protocol that exploits “Quadtree-
based network space partitioning” to provide more efficient
routing among multiple mobile stimuli and sink nodes has
been proposed. Simulation results show that in that work
significantly reduces average energy consumption while
maintaining comparably higher data delivery ratio [28]. A
novel computational paradigm for detection of blood vessels
in fundus images based on RGB components and quadtree
decomposition also has been proposed. The proposed
algorithm employs median filtering, quadtree decomposition,
 World Academy of Science, Engineering and Technology 22 2008
post filtration of detected edges, and morphological
reconstruction on retinal images [29]. A distributed quadtree
index that adapts the MX-CIF quadtree was described that
enables more powerful accesses to data in P2P networks. This
index has been implemented for various prototype P2P
applications [30].
The coordinate space (for the layer where all geometric
objects are located) is issued to a procedure called tessellation
in the linear quadtree indexing scheme of oracle spatial 10g,
which delineates elite and extensive cover tiles for all of pile
up geometry. Tessellation is done by crumbling the coordinate
space in a standard hierarchical mode. The range of
coordinates, the coordinate space, is viewed as a rectangle. At
the first level of decomposition, the rectangle is divided into
halves along each coordinate dimension, producing four tiles.
Each tile that cooperates with the geometry being tessellated is
further decomposed into four tiles. This procedure continues
until some termination criteria, such as size of the tiles or the
maximum number of tiles to cover the geometry is met. The
results of the tessellation process on geometry are stored in a
table, referred to as the SDOINDEX table [17].
Spatial quadtree indexing utilizes fixed-size tiles to cover
geometry. Fixed-size tiles are managed by tile resolution. If
the resolution is the single controlling reason, then tessellation
terminates when the coordinate space has been crumbled an
exact number of times. Consequently, each tile is of a fixed
amount and form. Fixed-size tile resolution is managed by a
user-selectable keyword named sdo_level. Smaller fixed-size
tiles give better geometry approximations.
IV. CONSIDERATION OF CHOOSING R- TREE OR QUADTREE
SPATIAL DATA INDEXING IN ONE SINGLE SPATIAL
DATABASE SYSTEM
R-tree and Quadtree indexes that use extensible framework
are the best spatial data indexing method from existing spatial
indexing method for indexing low-dimensional spatial data.
Oracle spatial users can choose between one of two spatial
indexes for indexing low-dimensional spatial data: a linear
Quadtree [31] or an R-tree [32, 33, 34, 35, 36, 37]. For
processing the queries, R-tree approach may be more efficient
due to better maintenance of spatial immediacy but may be
slow in updates or index creation and implements its own
concurrency protocols on top of table-level concurrency
mechanisms. Since R-tree is built logically as a tree and
physically using tables inside the database and the search
involves recursive SQL for traversing the tree from root to
relevant leaves. The linear Quadtree give results in simpler
index creation, faster updates and inheriting of builtin B-tree
concurrency control protocols. Since these indexes computes
tile approximations for geometries and uses existing Btree
indexes for performing spatial search and other DML
operations.
Oracle spatial provides spatial indexing to speed up access
and operation spatial data. That spatial index can be an R-tree
or Quadtree index, or both. Every type of indexes can give
good quality outcome in different situations. R-tree and
Quadtree index also can be used together at the same time in
one geometry column. To choose the appropriate indexing,
252
analysis of the data and operations indexing must be made in
advanced. The default of spatial index in oracle is an R-tree
which the dimension of the data is two. If the spatial index has
been built in two dimensions of data or more, only one built-
up operator (SDO_FILTER, the primary filter or index only in
the query) can be used in this situation. This operator
considers in all of dimension. Here we explore some of
comparisons between R-tree and Quadtree spatial data
indexing.
The method to find the interiors of convex polygons is
different and can’t be applied to concave geometries [39].
Different technique should be used in the concave geometries
to get the interior estimations for concave query windows.
One of the methods is to break up recursively the MBR
(Minimum Bounding Rectangle) of the concave geometry into
quadrants four times. This idea also can be applied to carry
out a level four tilling via the MBR as the tilling domain. Then
the geometries are recognized along with the tiles that wrap
the geometry. Aspirant MBRs are also tiled and the tiles for
the aspirant MBR are seek out along with the core tiles for the
concave query window. Current outcome [39] demonstrates
that this two-sided method for applying the intermediary filter
in R-trees attains extensive enhancements in query
performance. The core area detained using this method could
be larger than that for some Quadtree indexes assembled using
a tiling level of 14 since the tiling domain in that case is the
entire data space while the tiling domain for the R-tree is
simply the MBR of the concave geometry. Since this situation
happened, then the approximation of geometries cannot be
fine-tuned.
Quadtree indexing in Oracle Spatial database use two
strategies for index the geographical data. The first one is user
choose and specified the tilling level of Quadtree indexing
method. The second one is approximate every geometry data
thus the covering tiles can be minimize. At the appropriate
stage the tile-code associating each tile using z-ordering.
Tessellation process has a responsibility to divide the tiles for
geometry into internal and external part based on whether or
not they are fully interior to the geometry. The spatial_index
table receives all the wrapping tiles for geometry data beside
geometry rowid. The geometry data as a result of this method
may have numerous rows in the spatial_index table where
every dissimilar tile-code, rowed, and internal or external
status of the geometry stored in a row. Consequently geometry
approximation can be fine-tuned by setting the tilling level
and number of tiles. The last stage is speeding up queries for
construct B-tree index on the tile_code, rowed, and status.
R-tree has easier method in creating the index and tuning
rather than Quadtree spatial index creation. Because in Oracle,
R-tree will create the index by default without identify any
parameter. Moreover, setting the appropriate tuning parameter
values of Quadtree can affect performance significantly. In
order to optimize performance, Quadtrees necessitate to be
fine-tuned by choosing a fitting tiling level. The storage and
the index creation costs rise as the tiling level increases.
Quadtree require tilling level more than R-tree, as a result the
storage that is needed to create the Quadtree index is bigger
than using R-tree index.
 World Academy of Science, Engineering and Technology 22 2008
To recognize the nearest neighbors for a geometry,
SDO_NN (nearest neighbor) queries utilize the spatial index.
The parameter for tuning the parameter and may affect
query’s performance utilize SDO_BATCH_SIZE. The
SDO_BATCH_SIZE only can be used if SDO_NN utilize an
R-tree index to execute the procedure. Additionally this
keyword can’t be used when Quadtree index combined. The
value for this keyword will be affecting the performance of
nearest neighbor queries.
In Oracle Spatial, R-trees preserve their rational tree
formation and are executed as a table where each node of the
R-tree communicates to a row in the table and a child pointer
in the R-tree communicates to the row id of child row in the
same table. The root rowed (pointer) of the R-tree is stored in
the metadata for the index and allows routing from the root of
the R-tree to the leaf nodes. Leaf nodes of the R-tree
accumulate one MBR for each geometry data together with
the geometry row id. Queries and updates attain the root of the
R-tree and navigate down the tree to the leaves.
More details on this implementation can be obtained from
[38]. Note that each time an R-tree node is called; the
matching row from the spatial_index is chosen using a SQL
declaration internally. This means query and update
processing in R-trees occupies processing of more recursive
SQL declarations than in the case of Quadtrees. As a result,
some DML processes particularly update are likely to be more
costly. And the approximation of geometries cannot be fine-
tuned.
In the case of Quadtrees, internal tiles for data geometries
are recognized at index creation time. Every geometry tiles are
labeled with a "status" representing whether they are internal
tiles or external tiles. The status is set to boundary when any
element of the boundary of geometry touches a tile.
Otherwise, if the tile is totally inside the geometry, the status
is set to "interior". These status tags along with the tile-codes
are evaluated with those of the query tiles at query time to
execute the transitional filtering. Note that such labeling of the
tiles that cover query and data geometries inflicts very little
transparency (one byte per row) on storage space, and does
not change query or update performance.
A spatial R-tree index can index spatial data until four
dimensions unlike the Quadtree that only can index only two
dimensions data. Each geometry data approximated with R-
tree index by a particular rectangle that minimally attaches the
geometry using MBR. R-tree index consists of a hierarchical
index on the MBRs of the geometries for a sheet of geometries
in the layer. In the linear quadtree indexing design, the
coordinate space is focused to a procedure called tessellation,
which describes restrictedly and extensively cover tiles for
every stored geometry data. The coordinate space in a
standard hierarchical approach is tessellated by
decomposition. At the first level of disintegration, the
rectangle is divided into halves along each coordinate
dimension creating four tiles. Each tile that interrelates with
the geometry being tessellated is further decomposed into four
tiles. This procedure keeps on until some termination
condition, such as size of the tiles or the highest quantity of
tiles to cover the geometry, is met.
253
Linear referencing is an ordinary and suitable means to
correlate attributes or events to locations or portions of a
linear element. It has been extensively used in transportation
applications (such as for highways, railroads, and transit
routes) and efficacies applications (such as for gas and oil
pipelines). The major benefit of linear referencing is its
potentiality of locating attributes and events along a linear
element with only one constraint (usually known as quantify)
instead of two (such as latitude/longitude or x/y in Cartesian
space). Sections of a linear feature can be referenced and
generated dynamically by signifying the start and end
locations along the element without explicitly storing them. If
LRS (Linear Referencing System) data is indexed using a
spatial quadtree index, only the first two dimensions are
indexed; the measure dimension and its values are not
indexed.
SDO_WITHIN_DISTANCE utilizes the spatial index to
recognize the set of spatial objects that are within some
particular distance of a given object (such as an area of
interest or point of interest). Distance between two extensive
objects (nonpoint objects such as lines and polygons) is
described as the minimum distance between these two objects.
If this operator is used with geodetic data, the data must be
indexed with an R-tree spatial index. If this operator is used
with geodetic data and if the R-tree spatial index is created
with 'geodetic=false' recognized, we cannot use the part
parameter.
Geodetic data consists of geometries that have geodetic
SDO_SRID values, reflecting the actuality that they are stand
on a geodetic coordinate system (such as using longitude and
latitude) as contrast to a flat or estimated plane coordinate
scheme. Thus indexing geodetic data will improve
significantly spatial features. However, quadtree indexes
cannot be geodetic indexes. If we generate a non-geodetic
index on geodetic data, we cannot apply the part parameter
with the SDO_WITHIN_DISTANCE operator.
Distance, region, and angular units are completely
supported location data to be stored in the spatial database
using the whole-earth geodetic form which guarantees
capacities across the earth’s exterior will be very precise.
Location data can be stored in the database using the whole-
earth geodetic model An R-tree index is required for a whole-
Earth index. In another hand a quadtree index cannot be used
for a whole-Earth index.
To summarize all of those comparisons between R-tree and
Quadtree spatial data indexing, below is the table for
consideration to choose R-tree or Quadtree spatial data
indexing in one spatial database system.
 World Academy of Science, Engineering and Technology 22 2008

TABLE I productive in on-line work environment [41]. All of those

C HOOSING R-T REE OR QUADTREE I NDEXING [17]
functionalities of Oracle Application Server make GIS
No R-Tree Indexing Quadtree Indexing
application server to manage and organize Mobile GIS
1 The approximation of The approximation of
geometries cannot be fine- geometries can be fine-
application that run in web base.
tuned. (Spatial uses the tuned by setting the tiling There are some important tasks that we need to set and
minimum bounding level and number of tiles. configure after install Oracle Application Server and
rectangles.) MapViewer web application. Firstly we need to run Oracle
2 Index creation and tuning Tuning is more complex, Application Server instances in the script. Secondly configure
are easier. and setting the appropriate Oracle Application Server clusters using Oracle Process
tuning parameter values Manager and Notification Server (OPMN) commands. Thirdly
can affect performance enable and configure Secure Socket Layer (SSL), if we plan to
significantly. use Oracle Drive with Oracle Content DB and we are not
3 Less storage is required. More storage is required.
using Oracle Internet Directory as a user repository.
4 If your application If your application
In the implementation process using Oracle Application
workload includes nearest- workload includes nearest-
neighbor queries (SDO_NN neighbor queries Server MapViewer we used J2EE application that already
operator), R-tree indexes (SDO_NN operator), deployed to a J2EE container. In the rendering engine (Java
are faster, and you can use quadtree indexes are library) named SDOVIS we render cartographic USA Base
the sdo_batch_size slower, and you cannot Map. We used XML and AJAX-based JavaScript to store data
keyword. use the sdo_batch_size that will be retrieving for USA Base Map MapViewer web
keyword. application without directly connect to the spatial database. A
5 Heavy update activity to Heavy update activity graphical Map builder also used to create map symbols, define
the spatial column may does not affect the spatial data rendering rules, and create and edit USA Base
decrease the R-tree index performance of a quadtree Map objects.
performance until the index index.
is rebuilt.
Creating one of the geographical maps should be started
6 You can index up to four You can index only two from creating the XML transform. Here is one of the example
dimensions. dimensions. If LRS USA Base Map XML transform code to create XML from
(Linear Referencing MAP_ADMIN_AREA2_US table.
System) data is indexed
using a spatial quadtree XML Transform Code:
index, only the first two <?xml version="1.0" standalone="yes"?>
dimensions are indexed; <info_request datasource="adci" format="strict">
the measure dimension SELECT * FROM map_admin_area2_us
and its values are not
</info_request>
indexed.
7 An R-tree index is A quadtree index is not
recommended for indexing recommended for Result:
geodetic data if Indexing geodetic data if <ROWSET>
SDO_WITHIN_DISTANC SDO_WITHIN_DISTAN −
E queries will be used on it. CE queries will be used on <ROW num="1">
it. <POLYGON_ID>1717058563</POLYGON_ID>
8 An R-tree index is required A quadtree index cannot <AREA_ID>21013672</AREA_ID>
for a whole-Earth index. be used for a whole-Earth <POLYGON_NAME>VIRGINIA</POLYGON_NA
index. ME>
<NAME_LANGCODE>ENG</NAME_LANGC ODE>
V. MAPVIEWER AND APPLICATION SERVER OVERVIEW <FEATURE_TYPE>STATE</FEATURE_TYPE >
WITH THE I MPLEMENTATION <GEOMETRY>null</GEOMETRY>
This research uses Oracle Application Server 10g as the <PARTITION_ID>1</PARTITION_ID>
application server and MapViewer as web base application for </ROW>
our implementation. Oracle Application Server 10g that in </ROWSET>
mobile computing stands for grid consist of a combined, set-
stand software stage. In the main platform, Oracle Application Here is one of the source codes to display USA Base Map
Server consist of and OC4J (OracleAS Containers for J2EE) in MapViewer web base application with the main position in
and Oracle HTTP Server (built by Apache). This application Washington.
server deploys by J2EE programmable language. This …
application server turn into the starter platform planed for grid var mapview;
computing with full maintain for Service-Oriented
Architecture (SOA). It also has built in a middle tier server function showMap()
infrastructure. Additionally, it has a feature that can help {
construct our application to become more effective and

254
 World Academy of Science, Engineering and Technology 22 2008

var baseURL = <geometricProperty typeName="center">

"http://"+document.location.host <Point>
+"/mapviewer"; <coordinates>-77.0425, 38.87285</coordinates>
var mapCenterLon = -77.0425; </Point>
var mapCenterLat = 38.87285; </geometricProperty>
var mapZoom = 4; </geoFeature>
var mpoint = </center>
MVSdoGeometry.createPoint </map_request>
(mapCenterLon,mapCenterLat, 8307);
mapview = new MVMapView VI. PERFORMANCE EVALUATION
(document.getElementById ("map"), baseURL); The model used for testing and evaluation in this research
mapview.addBaseMapLayer (new has been simplified with the following assumptions [6]:
MVBaseMap ("adci.us_base_map")); 1. Time is one-dimensional and linearly ordered;
mapview.setCenter(mpoint); 2. Oracle 10g spatial databases are used to develop the GIS
mapview.setZoomLevel(mapZoom); model;
mapview.addNavigationPanel("EAST"); 3. Connection to the server and data retrieval via an active
TCP/IP connection using the Hyper Text Transfer Protocol
mapview.display(); (HTTP) utilized as data transfer protocol; the Mobile user
} provides an initial position.
… Simplified mobile GIS equipment has demonstrated how
This is the Map result from that source code. process retrieval and transferring data works. A mobile user
access data’s from server that uses Oracle 10g Spatial as a
database to show geographical map. We characterize the
performance of this function in the following way. The
average response time from the mobile client is measured as
the time spent (in seconds) from the moment the query is
issued to the moment the results of the query are generated.
The hardware settings are summarized in Table II.
TABLE II
S ETTING OF THE EXPERIMENTS
Web Server & CPU Intel Pentium Core
GIS Server: 2 Duo
RAM 2 GHz
Web Server IIS
GIS Server Oracle Application
Server &
MapViewer
Database Oracle 10g Spatial
Fig. 3 USA Base Map Basic Flow of Action with Oracle Application
Mobile Device Model Tablet PC
Server MapViewer
Equipment:
Here is the XML code to produce that one of US BASE OS Windows XP
Tablet PC
MAP.
Web Browser Internet Explorer
<?xml version="1.0" standalone="yes"?>
The result for R-tree spatial index is depicted in Fig. 4,
<map_request
which is plotted as a two-dimensional graph to illustrate the
title="USA Base Map"
results of average response time in Mobile GIS application.
basemap="us_base_map"
The line with blue color corresponds to the time measurement
datasource = "adci"
from the data that use an R-tree Spatial Indexing, while the
width="640"
line with red color represents the time measurement from the
height="480"
data that doesn’t use Spatial Indexing. We observe that the
bgcolor="#a6cae0"
time saved increases as the database size expands. For the
antialiase="false"
number of records, over 30,000, the time required in the data
format="PNG_STREAM">
that use an R-tree Spatial Indexing is almost two times more
<center size="0.15">
than the data that doesn’t use Spatial Indexing. It is noted,
<geoFeature render_style="m.star"
however, that this is related to the actual time of the query,
radius="1600,4800"
since the database volume is varied as time changes. The
label="A Place"
response time is slightly different to the result shown in Fig. 4.
text_style="t.Street Name" >
Nevertheless, we can easily discover that the time response on

255
 World Academy of Science, Engineering and Technology 22 2008

the data that use an R-tree Spatial Indexing is faster than the TABLE III
R- TREE AND Q UADTREE E XPERIMENTS R ESULT
data that doesn’t use Spatial Indexing.
Number No Index Quadtree No Index R-tree
of (Seconds) Index (Seconds) Index
Records - (Seconds) – (Seconds)
Quadtree R-tree
1000 1.4 1.2 1.4 1.3
5000 1.6 1.6 1.6 1.5
10000 2.5 2.4 2.5 2.3
15000 3 2.8 3 2.7
20000 3.5 3.5 3.5 3
25000 5 3.9 5 3.6
30000 6 4.3 6 4
35000 8 4.6 8 4.4

Fig. 4 R-tree Average Response Time
The result for Quadtree spatial index is depicted in Fig. 6,
which is plotted the results of average response time in Mobile
GIS application. The line with blue color corresponds to the
time measurement from the data that use a quadtree Spatial
Indexing, while the line with red color represents the time
measurement from the data that doesn’t use Spatial Indexing.
We observe that the time saved increases as the database size
expands. For the number of records, over 30,000, the time
required in the data that use a quadtree Spatial Indexing is
almost two times more than the data that doesn’t use Spatial
Indexing. It is noted, however, that this is related to the actual
time of the query, since the database volume is varied as time
changes. The response time is slightly different to the result
shown in figure (Fig. 5). Nevertheless, we can easily discover
that the time response on the data that use a quadtree Spatial
Indexing is faster than the data that doesn’t use Spatial
Indexing.
VII. CONCLUSION
Database tuning is an important step to solve system
bottleneck problem. The use of indexes is an efficient ways of
retrieving data from spatial database. R-tree and Quadtree
spatial index can be use as spatial database tuning methods to
enhance Mobile GIS performance. MapViewer also one of
web base applications that suitable, fast, and has a good
feature for mobile GIS applications to show geographical
map. Oracle Application Server and Oracle Spatial 10g have
function for web application server and recommended place to
store and retrieve spatial data. The goals are to fine-tune the
original database for mobile users with a limited
communication bandwidth, to improve the response time, and
show geographical map with a good feature. Further research
will develop and test this combination of R-tree and Quadtree
spatial data indexing in one spatial database system from those
of the considerations.
ACKNOWLEDGMENT
The authors would like to thank Ministry of Science,
Technology and Innovation Malaysia (MOSTI) and Universiti
Teknologi Malaysia (UTM) for their financial support under
e-science fund vot no. 79233.

Fig. 5 Quadtree Average Response Time
Below is the table (Table III) to make it clear to see the
result.
REFERENCES
[1] Rajinder, S. N. (2004). Cartographic visualisation for mobile
application. Master’s thesis, ITC/IIRS.
[2] Peng, Z. R. & Tsou, M. H. (2003). Internet GIS: Distributed Geographic
Information Services for the Internet and Wireless Network. John Wiley
and Sons Inc.
[3] Mensah, E. (2007). Designing a Prototype Mobile GIS to Support
Cadastral Data Collection in Ghana, 44.
[4] Vckovski, A. (1999). Interoperability and spacial information theory.
Interoperating Geographic Information Systems.
[5] Kraak, M. J. (2002). Current trends in visualisation of geographic data
with special reference to cartography. Invited paper: In Proceedings of
the XXIIth INCA Congress 2002, Indian National Cartographic
Association: Convergence of Imagery Information and Maps, volume
22, 319–324.
[6] Chen, K., & Shi, W. (2002), A Study of Dynamic Database in Mobile
GIS.
[7] Guttman, A. (1984). R-Trees: A Dynamic Index Structure for Spatial
Searching. Proc. 1984 ACM SIGMOD International Conference on
Management of Data, pp. 47-57. ISBN 0-89791-128-8.
[8] Zhu, Q., Gong, J., Zhang, J. (2007). An efficient 3D R-tree spatial index
method for virtual geographic environments. ISPRS Journal of

256
 World Academy of Science, Engineering and Technology 22 2008
Photogrammetry and Remote Sensing, Volume 62, Issue 3, August
2007, pp 217-224.
[9] Lee, T., Moon, B., Lee, S. (2006). Bulk insertion for R-trees by seeded
clustering. Data & Knowledge Engineering, Volume 59, Issue 1,
October 2006, pp 86-106.
[10] Lee, M. L., Hsu, W., Jensen, C. S., Cui, B., Teo, K. L. (2003).
Supporting Frequent Updates in R-Trees: A Bottom-Up Approach.
Proceedings 2003 VLDB Conference, 2003, pp 608-619.
[11] An, N., Kothuri, R. K. V., Ravada, S. (2003). Improving Performance
with Bulk-Inserts in Oracle R-Trees. Proceedings 2003 VLDB
Conference, 2003, pp 948-951.
[12] Chan, E. P. F., & Chow, K. K. W. (2002). On multi-scale display of
geometric objects. Data & Knowledge Engineering, Volume 40, Issue 1,
January 2002, pp 91-119.
[13] Yun, J. K., Kim, D. O., Hong, D. S., Kim, M. H., Han, K.J. (2006). A
real-time mobile GIS based on the HBR-tree next term for location
based services. Computers & Industrial Engineering, Volume 51, Issue
1, September 2006, pp 58-71.
[14] Francis, D. H., Madria, S., Sabharwal, C. (2008). A scalable constraint-
based Q-hash indexing for moving objects. Information Sciences,
Volume 178, Issue 6, 15 March 2008, pp 1442-1460.
[15] Liu, C. M., & Fu, S. Y. (2008). Effective protocols for kNN search on
broadcast multi-dimensional index trees. Information Systems, Volume
33, Issue 1, March 2008, pp 18-35.
[16] Finkel, R., & Bentley, J. L. (1974). Quad Trees: A Data Structure for
Retrieval on Composite Keys. Acta Informatica 4 (1): 1-9.
[17] Oracle Spatial 10g White Paper (2006). Oracle Spatial Quadtree
Indexing, 10g Release 1 (10.1).
[18] Chang, C. Y., Maciejewski, A. A., Balakrishnan, V., Roberts, R. G.,
Saitwal, K. (2006). Quadtree-based eigen decomposition for pose
estimation in the presence of occlusion and background clutter.
[19] Geller, S., Talke, J., Krafczyk, M. (2007). Lattice-Boltzmann Method on
Quadtree-Type Grids for Fluid Structure Interaction. Fluid-Structure
Interaction, 270-293.
[20] Tzouramanis, T., Vassilakopoulos, M., Manolopoulos, Y. (2000).
Multiversion Linear Quadtree for Spatio-Temporal Data. Current Issues
in Databases and Information Systems, 279-292.
[21] Zhang, L. & Xi, L. F. (2007). A Novel Fractal Image Coding Based on
Quadtree Partition of the Adaptive Threshold Value. Theoretical
Advances and Applications of Fuzzy Logic and Soft Computing, 504-
512.
[22] Varas, J. (2007). Mobile Robot Path Planning Among Weighted Regions
Using Quadtree Representations. Computer Aided Systems Theory -
EUROCAST’99, 239-249.
[23] Cheng, S. W. & Lee, K. H. (2008). Quadtree Decomposition, Steiner
Triangulation, and Ray Shooting. Algorithms and Computation, 368-
377.
[24] Grbovic, J. P., Fagg, G. E., Angskun, T., Bosilca, G., Dongarra, J. J.
(2006). MPI Collective Algorithm Selection and Quadtree Encoding.
Recent Advances in Parallel Virtual Machine and Message Passing
Interface, 40-48.
[25] Varas, J. (2007). Mobile Robot Path Planning Among Weighted Regions
Using Quadtree Representations. Computer Aided Systems Theory -
EUROCAST’99, 239-249.
[26] Mir, Z. H. & Ko, Y. B. (2006). A Quadtree-Based Data Dissemination
Protocol for Wireless Sensor Networks with Mobile Sinks. Personal
Wireless Communications, 447-458.
[27] Samet, R. & Ozsavas, E. (2007). Optimization of Quadtree
Triangulation for Terrain Models. Advanced Concepts for Intelligent
Vision Systems, 48-59.
[28] Mir, Z. H. & Ko, Y. B. (2007). A quadtree-based hierarchical data
dissemination for mobile sensor networks. Telecommunication
Systems, 117-128.
[29] Reza, A. W., Eswaran, C., Hati, S. (2007). Diabetic Retinopathy: A
Quadtree Based Blood Vessel Detection Algorithm Using RGB
Components in Fundus Images. Journal of Medical Systems, 147-155.
[30] Tanin, E., Harwood, A., Samet, H. (2006). Using a distributed quadtree
index in peer-to-peer networks. The VLDB Journal, The International
Journal on Very Large Data Bases, 165-178.
[31] H. Samet (1989). The design and analysis of spatial data structures.
Addison-Wesley Publishing Co..
257
[32] A. Guttman (1984). R-trees: A dynamic index structure for spatial
searching. Proc. A CM SIGMOD Int. Conf. on Management of Data,
pages 47-57.
[33] T. Sellis, N. Roussopoulos, and C. Faloutsos (1988). The r+-tree: A
dynamic index for multi-dimensional objects. Procdf the Int. Conf. on
Very Large Data Bases, 13:507-518.
[34] N. Beckmann, H. Kriegel, R. Schneider, B. Seeger (1990). The R* tree:
An efficient and robust access method for points and rectangles. In Proc.
ACM SIGMOD Int. Conf. on Management of Data, pages 322-331.
[35] S. Berchtold, D. A. Keim, and H. P. Kreigel (1996). The X-tree: An
index structure for high dimensional data. Procff the Int. Conf. on Very
Large Data Bases.
[36] S. T. Leutenegger, M. A. Lopez, J. M. Edgington (1997). STR: A simple
and efficient algorithm for R-tree packing. In Proe. Int. Conf. on Data
Engineering.
[37] V. Gaede & O. Gunther (1998). Multidimensional access methods. ACM
Computing Surveys, 30(2).
[38] K. V. Ravi Kanth, Siva Ravada, J. Sharma, J. Banerjee (1999). Indexing
medium-dimensionality data in oracle. In Proc. ACM SIGMOD Int.
Conf. on Management of Data.
[39] K. V. Ravi Kanth & Siva Ravada (2001). Efficient processing of large
spatial queries using interior approximations. In Symposium on Spatial
and Temporal Databases (SSTD).
[40] Oracle Application Server 10g White Paper (2007). Oracle Application
Server 10g.
Maruto Masserie Sardadi obtained his B.Sc. degree
in Computer Science from the University of
Indonesia (UI), Indonesia in 2004. All of his working
experienced is in IT area, some in oil industry. He is
currently a Computer Science M.Sc. research student
in University Technology of Malaysia (UTM). His
research interests include database, GIS, and
petroleum.
Mohd Shafry Mohd Rahim obtained his B.Sc.
degree in Computer Education, 1999 and M.Sc.
degree in Computer Graphics, 2002 from University
Technology of Malaysia (UTM). He is currently a
Computer Science PhD. research student in
University Putra of Malaysia (UPM) and UTM
Computer Science lecturer. His research interests
include GIS and spatiotemporal database.
Zahabidin Jupri obtained his B.Sc. degree in
Mathematics Education from University of Malaya
(UM) and M.Sc. degree in Info. Mgmt. from
Sheffield, UK. He is currently UTM Computer
Science lecturer.
Daut Daman, obtained his B.Sc. degree in Math.
Computer Science, 1979 from University Sains of
Malaysia (USM) and M.Sc. degree in Applied
Computing, 1984 from Cranfield Institute of
Technology, UK. He is currently a Head Department
of Computer Graphics and Multimedia Faculty of
Computer Science UTM.